We’ve been discussing the nature of student interactions with the compiler. In particular, we know (informally) there are different types of students; we’ve seen them at work (or hacking), and they behave differently. Can we classify this behavior; that is, can we use the student’s pattern of interaction with the compiler as a predictor or a way to classify them early, so we might assist where necessary?
This is not at all about finding plagarism. I don’t think the problem is the same. In this case, I expect the code to be quite similar between compiles, and I am looking for metrics to help measure how different any two subsequent compiles are.
Tonight I’m going to run some algorithms on some code by hand; I’ll use STRING-EDIT (which I’ll actually use the Perl String::Approx module for), and will do the analysis of the parse trees of a few programs by hand. Then, I’ll have some numbers to kick around for a few, simple, but likely types of student changes to code.
I hope this whole sidetrack is productive…